Modelagem de dados com medidas repetidas via equações de estimação generalizadas

Abstract

In many situations of intereset it is commom to have more than one observation for experimental unit, thus generating the experiments with repeated measures. In the modeling of such experiments is necessary consider and model the intra-unit dependency structure. The first modeling proposals were based on normality assumption, but do not always present a good alternative. That being said, an flexibilization alternative, is to make use of Generalized Estimating Equations (GEE’s). These models use properties of estimating functions to build estimating equations which incorporate an correlation structure. Such equations, initially, were obtained for Generalized Linear Models, whose we will deal with here. In this work the theory of estimating functions and the constructions of GEE’s will be presented, as well as techniques for model selections and working correlation matrix selection, statistics to test linear hypothesis of interest with respect to regression parameters, as well as diagnostic techniques of global and local influence. The importance and necessity of modeling using GEE’s will be presented and discussed through practical examples.